Managing Connections

sqlhelper can maintain one or many database connections internally. This article describes how to configure and create connections, and the functions for inspecting and accessing them.

Creating connections

You can define database connections in config files which are read and processed by the connect() function. For example, if your config file is called my_db_config.yml, you might run:

library(sqlhelper)

connect("examples/sqlhelper_db_conf.yml")

Config files

A config file can contain one or more connection definitions. The above example defines two simple connections:

readLines("examples/sqlhelper_db_conf.yml") |>
  writeLines()
#> #### sqlhelper_db_conf.yml ####
#> 
#> simple_sqlite:
#>   driver_type: sqlite
#>   description: "A simple connection to an in-memory database"
#>   connection:
#>     Server: ":memory:"
#> 
#> pool_sqlite:
#>   driver_type: sqlite
#>   pool: yes
#>   description: "A pooled connection to an in-memory database"
#>   connection:
#>     Server: ":memory:"

Connection definitions are YAML. The following YAML chunk defines a connection called ‘dap’.

dap:
  driver_type: sqlserver
  pool: yes
  description: >
    Databases managed by ADD teams on the Data and Analytics Platform
  connection:
    Driver: "{ODBC Driver 17 for SQL Server}"
    Server: "Dap-sql01"
    Trusted_Connection: "yes"

The first line is a name for this connection; subsequent lines contain information about the connection.

The driver_type line defines the brand of database, and hence the driver package that will be used. This parameter is required. Current options are:

The pool line determines whether a single connection is required (as returned by DBI::dbConnect()) or a pool of connections (as returned by pool::dbPool()). The latter is recommended for shiny applications. This parameter defaults to ‘no’, meaning a single connection will be provided.

The description line provides a short description of this connection. This parameter can be omitted.

The connection section contains several lines which provide the parameters required for the connection itself, to be passed to the constructor function as name=value pairs. For example, you will need to supply a Driver connection string if you are using an odbc driver. The server parameter will probably always be required but see DBI::dbConnect() for more details.

In summary, the elements needed for every connection are:

  1. The connection name, and
  2. the driver_type, and
  3. a connection section containing a Server element

The others may be needed for some connections or may be omitted entirely. The function config_examples() provides several example configurations.

The config search path

By default, the connect() function reads not only the config file named by its config_filename parameter, but also any config files found on a search path that includes both the user’s and the site’s config directories.

If you use the same database connections often, you can place a config file called sqlhelper_db_config.yml in the directory named by rappdirs::user_config_dir(). Similarly, if you are a site admin you can provide your users with pre-configured connections by placing a config file of the same name in the directory named by rappdirs::site_config_dir(). In either of these cases, no config filename is needed and connect() can be called without arguments. Nonetheless, in the case where you require additional connections to those provided by user- and site-wide configurations, you may provide a config_filename argument.

Suppressing the config search path

If only one config file is wanted, the config search path may be suppressed by supplying the exclusive = TRUE argument.

If exclusive = TRUE, then connect() will check whether the config_filename parameter is ‘user’ or ‘site’. If config_filename is one of those, the user- or site-wide config file will be sought; if it is not, it will be treated as the name of a config file, and that file will be sought.

File precedence, inheritance, and conflicts across config files

If exclusive == FALSE, site-wide config files will be read first, then user-wide, then named files. If conflicting connection names or connection elements are encountered between config files, values from later files overwrite those from earlier. This mechanism operates element-wise; for example, if your site-wide config contains:

my_conn:
  server_type: mysql
  connection:
    Server: "organization_server_host"

but your named file contains:

my_conn:
  description: Private MySQL database
  connection:
    Server: "my_alternate_server_host"

then sqlhelper will attempt to create a connection described by:

my_conn:
  server_type: mysql
  description: Private MySQL database
  connection:
    Server: "my_alternate_server_host"

Browsing the connection cache

After connect() has been called, you may inspect the created connections with connection_info().

connection_info()
#> # A tibble: 2 x 7
#>   name          description                  live  default driver conn_str pool 
#>   <chr>         <chr>                        <lgl> <lgl>   <glue> <chr>    <lgl>
#> 1 pool_sqlite   A pooled connection to an i~ TRUE  FALSE   RSQLi~ <NA>     TRUE 
#> 2 simple_sqlite A simple connection to an i~ TRUE  TRUE    RSQLi~ <NA>     FALSE

Accessing individual connections

Connections may be accessed by name, with live_connection()

myconn <- live_connection("simple_sqlite")
myconn
#> <SQLiteConnection>
#>   Path: 
#>   Extensions: TRUE

The default connection

The first connection in the file with the highest precedence (i.e. the last one to be read) becomes the default default. This connection will be used whenever queries or files of SQL are executed (runqueries() or runfiles()) with without explicitly stating which connection to use.

You can check the default connection with connection_info() or get_default_conn_name() and change it with set_default_conn_name().

get_default_conn_name()
#> [1] "simple_sqlite"

set_default_conn_name("pool_sqlite")

get_default_conn_name()
#> [1] "pool_sqlite"

get_default_conn_name() |>
  live_connection()
#> <Pool> of SQLiteConnection objects
#>   Objects checked out: 0
#>   Available in pool: 1
#>   Max size: Inf
#>   Valid: TRUE

The last of these - combining get_default_conn_name() with live_connection() to obtain the default connection - is common enough to warrant a convenience function, default_conn().

Checking, closing and re-opening connections

You can check whether connections are live in two ways:

  1. connection_info() provides a ‘live’ field for interactive use
  2. The functions is_connected() and not_connected() are intended for programmatic use.

To close all connections and remove them from the internal cache, use disconnect().

Individual connections may be closed with DBI::dbDisconnect().

connect() closes any open connections before reading config files.

conn_name <- "simple_sqlite"

is_connected(conn_name)
#> [1] TRUE
connection_info(conn_name)$live
#> [1] TRUE

myconn <- live_connection(conn_name)
DBI::dbDisconnect(myconn)

connection_info()
#> # A tibble: 2 x 7
#>   name          description                  live  default driver conn_str pool 
#>   <chr>         <chr>                        <lgl> <lgl>   <glue> <chr>    <lgl>
#> 1 pool_sqlite   A pooled connection to an i~ TRUE  TRUE    RSQLi~ <NA>     TRUE 
#> 2 simple_sqlite A simple connection to an i~ FALSE FALSE   RSQLi~ <NA>     FALSE

if(not_connected(conn_name)){
  message(glue::glue("{conn_name} is not available, reconnecting..."))
  connect("examples/sqlhelper_db_conf.yml", exclusive = TRUE)
}
#> simple_sqlite is not available, reconnecting...

connection_info()
#> # A tibble: 2 x 7
#>   name          description                  live  default driver conn_str pool 
#>   <chr>         <chr>                        <lgl> <lgl>   <glue> <chr>    <lgl>
#> 1 pool_sqlite   A pooled connection to an i~ TRUE  FALSE   RSQLi~ <NA>     TRUE 
#> 2 simple_sqlite A simple connection to an i~ TRUE  TRUE    RSQLi~ <NA>     FALSE

disconnect()

connection_info()
#> NULL